Scientists have developed a pocket-sized AI brain, made possible by the integration of monkey neurons. This breakthrough, detailed in the journal Nature, showcases an AI model that efficiently replicates the visual system of the brain, reducing its size to a tiny fraction of its original form. The model, initially utilizing 60 million variables, was compressed to a version with just 10,000 variables, performing nearly as well. This achievement not only highlights the potential for more energy-efficient AI but also offers insights into the human brain's remarkable capabilities. The study, conducted in collaboration with researchers from Carnegie Mellon and Princeton Universities, focused on understanding the human visual system, which transforms light into recognizable objects. By compressing the AI model, scientists gained insights into the specialized functions of V4 neurons, which respond to specific shapes and patterns, such as arranged fruit in a supermarket. This discovery could lead to more efficient and human-like AI, potentially enhancing applications like self-driving cars. However, the AI still faces challenges in tasks like recognizing faces from various angles, suggesting the need for further advancements in AI technology.
